LBDNet inter-laboratory comparison at high doses of ionizing radiation using the dicentric plus caffeine assay

To assess the performance of the LBDNet laboratories in estimating dose over 5 Gy of ionizing radiation using the dicentric chromosome plus caffeine assay. Dose-response curve fitting: Peripheral blood was irradiated <i>in vitro</i> between 5 and 25 Gy. Then, the DC plus caffeine assay was carried out. Thirteen laboratories received and analyzed metaphase images. The linear dose-response curve was fitted for each laboratory. Dose estimation was performed analyzing coded metaphase images from three different irradiated samples (7.5, 15, 20 Gy) and using the fitted curve from every laboratory. The dose estimation accuracy was within the expected dose ranges. The 76.9%, 84.6% and 69.2% of the estimated doses fell into the ± 20% of the true radiation dose. The 92.3%, 92.3%, and 61.5% of the 95% of the confidence interval of the estimated doses included the true radiation dose. The trueness was 0.9%, 4.4% and 9.6%. The Coefficients of Variation of the estimated doses were 14.5%, 16.1% and 17.8%. Results from only one laboratory were deemed questionable for dose estimation, based on the <i>Z-score</i> derived from robust methods. The intercomparison study yielded satisfactory results; however, dose estimation accuracy tended to decrease, and variability between laboratory results increased as the dose level rose.

为评估LBDNet实验室采用双着丝粒染色体+咖啡因试验（dicentric chromosome plus caffeine assay）估算5Gy以上电离辐射剂量的性能，开展了本次剂量-响应曲线拟合研究：将外周血在5~25Gy范围内进行体外（in vitro）辐照，随后开展双着丝粒染色体+咖啡因试验。共有13家实验室接收并分析了中期分裂象图像，并为每家实验室拟合了线性剂量-响应曲线。剂量估算环节通过对三份不同辐照剂量样品（7.5、15、20Gy）的编码中期分裂象图像进行分析，并结合各实验室的拟合曲线完成。结果显示，剂量估算准确度处于预期范围内：分别有76.9%、84.6%与69.2%的估算剂量落在真实辐射剂量的±20%误差区间内；在估算剂量的95%置信区间中，分别有92.3%、92.3%与61.5%的区间包含了真实辐射剂量。对应三个剂量点的估算剂量正确度分别为0.9%、4.4%与9.6%，变异系数分别为14.5%、16.1%与17.8%。基于稳健方法推导的Z值（Z-score），仅有1家实验室的剂量估算结果被判定为存疑。本次实验室间比对研究整体结果令人满意，但随着辐射剂量水平升高，剂量估算准确度呈下降趋势，且实验室间结果的离散程度也随之增大。